Research On Detection Identification Algorithm Of Drogue Images And Flight Verification

Aer ial refueling technology is a powerful method to expand the combat radius and to prolong the flight time in the modern information wars,which is of great signif icance to enhance the combat capability of the Manned Aerial Vehicle(MAV)and Unmanned Aer ial Vehicle(UAV).However,current measure of manual air refueling suffers from great difficulties and dangers,as well as consumes much time.It will help to increase the efficiency of air refueling and decrease difficulties if Autonomous Aer ial Refueling(AAR)system can be established,which is an auxiliary to the refueling process.In this thesis,visual processing algorithms in AAR system based on vision technique are studied.Besides,experimental platform based on six-rotor UAV is built in order to test and verify algorithms.Firstly,methods of detecting the drogue used in air refueling are studied.Pretreatments are finished via image graying,threshold segmentation and morphological filter ing.After extracting feature contour,target will be captured by ellipse fitting based on least square method.Then how to detect the drogue when it is blocked is studied and strategies of detecting are designed separately according to different distances.Secondly,adaptive methods of tracking the drogue image are studied.A tracking method which uses simple enlargement frame is introduced.Moreover,image tracking strategy based on Meanshift algor ithm is discussed.A new method on the basis of adaptive kernel window is proposed to improve Meanshift algorithm and it is verified by experiments.Thirdly,semi-hardware AAR experiment platform which relies on six-rotor UAV is built.Construction of hardware can imitate real circumstance of aerial refueling and requirements of system.Software parts to realize communications in system,strategies of navigation and optimal control is designed.Program framework of multi thread is achieved simultaneously.Finally,algorithms of target detecting and tracking in dynamic flight condition are verified by using semi-hardware platform.Experiments to compare efficiencies between global detecting and local tracking,single thread and multi thread are carried out.The results reveal that vision-based AAR system in this thesis is feasible and vision algor ithms are effective.Detecting and tracking the drogue target in the circumstance of dynamic flight can be accomplished.Furthermore,simulation of docking in aerial refueling can be realized.